Social-based recommendation systems exploit the selections of friends to\ncombat the data sparsity on user preferences, and improve the recommendation\naccuracy of the collaborative filtering strategy. The main challenge is to\ncapture and weigh friends' preferences, as in practice they do necessarily\nmatch. In this paper, we propose a Neural Attention mechanism for Social\ncollaborative filtering, namely NAS. We design a neural architecture, to\ncarefully compute the non-linearity in friends' preferences by taking into\naccount the social latent effects of friends on user behavior. In addition, we\nintroduce a social behavioral attention mechanism to adaptively weigh the\ninfluence of friends on user preferences and consequently generate accurate\nrecommendations. Our experiments on publicly available datasets demonstrate the\neffectiveness of the proposed NAS model over other state-of-the-art methods.\nFurthermore, we study the effect of the proposed social behavioral attention\nmechanism and show that it is a key factor to our model's performance.\n